Design-based and model-based inference in survey sampling: appreciating the difference

Model-based ideas in finite-population sampling have received renewed discussion in recent years. Their relationship to the classical ideas in sampling theory do not appear to be universally well understood by samplers in applied disciplines such as forestry, and ecology more broadly. The two inferential paradigms are constrasted, and explanations are supplemented with examples of discrete as well as continuously distributed populations. The treatment of spatial structure is examined, also.